Spiking activity from neurophysiological experiments often exhibits dynamics beyond that driven by external stimulation, presumably reflecting the extensive recurrence of neural circuitry. Characterizing these dynamics may reveal important features of neural computation, particularly during internally-driven "cognitive" operations. For example, neurons in premotor cortex (PMd) are active during a "planning" period separating movement-target specification and a movement-initiation cue. Recent evidence suggests that PMd neural activity settles to a movement-specific state during this period. Can trial-to-trial variation in behavior be predicted from the dynamics of settling? Current methods to characterize recurrent neural dynamics on a trial-by-trial basis, and thus answer this and related questions, are limited. Standard methods average activity from different trials or different cells, and so cannot express variable dynamics. The proposed research will test the hypothesis that the dynamics underlying PMd plan activity can be described by a low-dimensional hidden non-linear dynamical systems (HNLDS) model, with underlying recurrent structure and stochastic point-process output. Such a model is capable of expressing rich dynamics, but the task of learning the model parameters from spike data is challenging. The proposed research will develop and validate algorithms for parameter estimation, and then characterize the dynamics in PMd data recorded from an electrode array while monkeys perform delayed-reach tasks. Single trial estimates of underlying dynamics can then be used to predict variation in details of reaching motor behavior. The proposed research program will directly inform cortically-controlled neural prosthesis research in our laboratory and elsewhere. Such motor and communication prostheses could dramatically improve the quality of life for patients with upper spinal cord injuries, amputations, ALS and other neuro-degenerative diseases. The proposed research program will increase our understanding of how PMd rapidly prepares movements, and thereby help increase the speed and accuracy of prosthetic systems.